This crash could happen at any time a RTP and RTCP buffer arrived
simultaneously in ssrcdemux.
The problem was that sticky-event arriving while the rtp and rtcp pads
were being set up could arrive just too late to be included in the initial
forwarding.
The fix checks if the stickies have been sent on the srcpad about to be
pushed on, and if not sends them. It also blocks any stickes from
being forwarded *prior* to this happening, to avoid them arriving on
the srcpad multiple times.
Since the test loops 1000 times, this will make running under valgrind
take forever, so use the RUNNING_ON_VALGRIND variable to detect we
are running under valgrind, and reduce the loop-count to 2 in that case.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-good/-/merge_requests/992>
When used for processing bundled media streams within rtpbin the rtpssrcdemux element may
receive bad RTP and RTCP packets, these should not be treated as a fatal error.
The property is useful against atacks when the sender changes SSRC for
every RTP packet. The property with the same name introduced in rtpbin
was not enough, because we still can end up with thousands of pads
allocated in rtpssrcdemux.
By passing `NULL` to `g_signal_new` instead of a marshaller, GLib will
actually internally optimize the signal (if the marshaller is available
in GLib itself) by also setting the valist marshaller. This makes the
signal emission a bit more performant than the regular marshalling,
which still needs to box into `GValue` and call libffi in case of a
generic marshaller.
Note that for custom marshallers, one would use
`g_signal_set_va_marshaller()` with the valist marshaller instead.
In this change we now protect the internal srcpads list using the
stream lock and limit usage of the internal stream lock to
preventing data flowing on the other src pad type while creating
and signalling the new pad.
This fixes a deadlock with RTPBin shutdown lock. These two locks would
end up being taken in two different order, which caused a deadlock. More
generally, we should not rely on a streamlock when handling out-of-band
data, so as a side effect, we should not take a stream lock when
iterating internal links.
This is an extra internal recurisve lock use to avoid having to take
both sink pad streams lock all the time. This patch renamed it
INTERLNAL_STREAM_LOCK/UNLOCK() to avoid confusion with possible upstream
GST_PAD API.
This reverts "6f3734c305 rtpssrcdemux: Only forward stick events while
holding the sinkpad stream lock" and actually hold on the internal
stream lock. This prevents in some needed case having a second
streaming thread poping in and messing up event ordering.
While forwarding serialized event, we use gst_pad_forward() function.
In the forward callback (GstPadForwardFunction) we always return
TRUE. Returning true there will stop the dispatching procedure. As a
side effect, only one events is receiving the events. This breaks
when sending EOS from the applicaiton, it also breaks the latency
tracer.
According to RFC 5506, reduce size packages can be sent, this
packages may not be compound, so we need to add support for
getting ssrc from other types of packages.
https://bugzilla.gnome.org/show_bug.cgi?id=750327
They are very confusing for people, and more often than not
also just not very accurate. Seeing 'last reviewed: 2005' in
your docs is not very confidence-inspiring. Let's just remove
those comments.
Otherwise we get a race where if the RTCP packet comes in first and while
it is added the pads, the segment event arrives on the RTP stream, the event
may be lost completely and never forwarded.
Block the RTP pad and associated RTCP pads while they are being
announced. This it to prevent a race where one is announced and
before the callback has connected it, the other one gets a buffer.
We can't use the "padlock" of ssrcdemux because it causes deadlocks.
This prevents a deadlock where something would try to push an event
through the SSRC demux from the callback, causing the pads to be iterated
and the lock taken.
